Using exposome-wide association studies (EWAS) t di f(EWAS) to discover causes of cancer

S.M. RappaportS.M. RappaportUniversity of California BerkeleyUniversity of California BerkeleyUniversity of California, BerkeleyUniversity of California, Berkeley

Research support from NIEHSResearch support from NIEHSResearch support from NIEHSResearch support from NIEHS

Some backgroundSome backgroundSome backgroundSome background•• About three fourths of all people die from About three fourths of all people die from

chronic diseases, mainly CVD and cancerchronic diseases, mainly CVD and cancer•• These diseases likely result from a combination These diseases likely result from a combination yy

of genetic (G) and environmental (E) factorsof genetic (G) and environmental (E) factors•• But how much of the risk can be attributed to GBut how much of the risk can be attributed to GBut how much of the risk can be attributed to G, But how much of the risk can be attributed to G,

E and E and GxEGxE??

2SM Rappaport

Explained variance of cancer incidenceExplained variance of cancer incidence(From structural equation modeling of the Swedish Family(From structural equation modeling of the Swedish Family--

Si G i Shared Childhood Non-shared

(From structural equation modeling of the Swedish Family(From structural equation modeling of the Swedish FamilyCancer database of 10M individuals born after 1934)Cancer database of 10M individuals born after 1934)

Site Genetic S a edenvironmental

C d oodenvironmental

o s a edenvironmental

Stomach 0.01 0.15 0.13 0.71Colon 0.13 0.12 0.06 0.69

Rectum 0.12 0.09 0.03 0.75Lung 0.08 0.09 0.04 0.79Lung 0.08 0.09 0.04 0.79Breast 0.25 0.09 0.06 0.60Cervix

(invasive) 0.22 0.00 0.03 0.75

Cervix (in situ) 0.13 0.00 0.13 0.74

Testis 0.25 0.00 0.17 0.58Kidney 0.08 0.08 0.06 0.78Bladder 0.07 0.12 0.04 0.77

Melanoma 0.21 0.02 0.08 0.69Nervous system 0.13 0.05 0.02 0.80

Thyroid 0.53 0.01 0.10 0.36Endocrine 0.28 0.03 0.11 0.58

Non-Hodgkin’s lymphoma 0.10 0.06 0.02 0.83

Leukemia 0.01 0.08 0.04 0.88Median 0.13 0.07 0.06 0.75

3SM Rappaport

K Czene, P. Lichtenstein and K Hemminki, Int J Cancer 2002, 99: 260-6

Attributable riskAttributable riskAttributable riskAttributable risk“The population attributable fraction (“The population attributable fraction (PAFPAF) can ) can be interpreted as be interpreted as the proportion of disease cases the proportion of disease cases over a specified time that would be prevented over a specified time that would be prevented following elimination of the exposuresfollowing elimination of the exposures, , assuming the exposures are causal.” assuming the exposures are causal.”

B B RockhillRockhill, B Newman and C Weinberg, AJPH, 1998, 88: 15, B Newman and C Weinberg, AJPH, 1998, 88: 15--19.19.

4SM Rappaport

Familial risks of cancerFamilial risks of cancer(From Swedish Family(From Swedish Family--Cancer database)Cancer database)

Si C iFamilial PAF

(%)Site Case pairs (%)Prostate 559 20.55*Breast 2784 10.61*

Colorectum 771 6.87Colorectum 771 6.87Endometrium 119 5.31*

Ovary 155 4.90*Lung 330 3.81

Over 22 sites the median PAF = 1.4 %

Thyroid 102 3.56Melanoma 382 2.74

Testis 63 2.71*Cervix 122 2 43Cervix 122 2.43Skin 75 2.35

Bladder 146 2.03All others < 2.00

5SM Rappaport K Hemminki and K Czene, CEBP 2002, 11: 1638-44

* PAF was doubled to reflect both parental lineages.

Environmental risks of cancerEnvironmental risks of cancerEnvironmental risks of cancerEnvironmental risks of cancerUrban air pollution

Contamin. injections Unsafe sex

Diet & exercise*

Indoor smoke

ject o s Unsafe sex

Alcohol

SmokingNot

attributedattributed (65%)

Attributable risks for cancer (worldwide, all tumor types, joint PAF=35%)

6

SM RappaportData from Ezzati et al., “Comparative Quantification of Mortality and Burden of Disease Attributable to Selected Risk Factors,” Global Burden of Disease and Risk Factors, Chapter 4, WHO, 2006.

Discovering causes of cancerDiscovering causes of cancerDiscovering causes of cancerDiscovering causes of cancer•• Cancer risks attributable to genetic factors Cancer risks attributable to genetic factors

(G) are typically small (1 (G) are typically small (1 –– 2%) 2%) ( ) yp y (( ) yp y ( ))•• Most cancers must be caused by nonMost cancers must be caused by non--

genetic factors (E) orgenetic factors (E) or GxEGxEgenetic factors (E) or genetic factors (E) or GxEGxEoo However, two thirds of attributable E risks However, two thirds of attributable E risks

h t b id tifi dh t b id tifi dhave not been identifiedhave not been identified•• What tools are available for identifying G, E What tools are available for identifying G, E y gy g

and and GxEGxE causes of cancer?causes of cancer?7

SM Rappaport

Human genotyping: major technology advancesSNPs per assay

1997 1

2001 10

2002 1,000

2004 50,000

2006 500,000

2007 1,000,000

2010 >>1,000,000

Genome-Wide Association Studies (GWAS) now possible with 2 000 20 000 samples (2 billion 20 billion genotypes)with 2,000-20,000 samples (2 billion - 20 billion genotypes)

Courtesy of E. Lander, MIT/Broad

Environmental factors in epidemiologyEnvironmental factors in epidemiologyEnvironmental factors in epidemiologyEnvironmental factors in epidemiology

Two thirds of studiesTwo thirds of studies relied upon subjects to assess their own exposures!

S. M. Rappaport9

B.K. Armstrong et al. Principles of Exposure Measurement in Epidemiology, Oxford Med. Pubs., 1992

Exposure assessment for Exposure assessment for ppcancer (2010)cancer (2010)

10SM Rappaport

DM Parkin, The fraction of cancer attributable to lifestyle andenvironmental factors in the UK in 2010, Brit J Cancer 105, S1-S5 (2011).

Finding unknown causes of cancerFinding unknown causes of cancer•• Elaboration of the G matrix with modern GWAS Elaboration of the G matrix with modern GWAS

has been stunningly comprehensive has been stunningly comprehensive oo but has explained relatively little cancer riskbut has explained relatively little cancer risk

•• Elaboration of the E matrix relies on Elaboration of the E matrix relies on questionnaires, geographic information and questionnaires, geographic information and targeted measurementstargeted measurementsggoo much as it did a century agomuch as it did a century ago

11SM Rappaport

The exposome The exposome –– promoting discovery promoting discovery of environmental causes of diseaseof environmental causes of disease

Christopher Wild defined the ‘exposome’, Christopher Wild defined the ‘exposome’, representing representing allall environmental exposures environmental exposures (including diet, lifestyle, and infections) from (including diet, lifestyle, and infections) from conception onwards, as a complement to the conception onwards, as a complement to the genome in studies of disease etiology.genome in studies of disease etiology.

Wild C P Cancer Epidemiol Biomarkers Prev 14 (8) 1847 1850 (2005)Wild, C.P., Cancer Epidemiol Biomarkers Prev 14 (8), 1847-1850 (2005).

12SM Rappaport

13SM Rappaport

Scientific citations to ‘exposome’ Scientific citations to ‘exposome’ pp(Google Scholar)(Google Scholar)

200

s

150

cita

tions

1st NAS k h

2nd NAS workshop

50

100

mbe

r of c

Wild CEBP(2005)

1st NAS workshop; Science Perspective

02004 2006 2008 2010 2012

Nu

2004 2006 2008 2010 2012Year

14SM Rappaport

Capturing exogenous and Capturing exogenous and endogenous exposuresendogenous exposuresendogenous exposuresendogenous exposures

The exposomeincludes all chemicals in the i t l h i linternal chemical environment

A. Schetter and C. Harris, PNAS, 2012, 109: 7955-6

15SM RappaportS.M. Rappaport and M.T. Smith, Science, 2010: 330, 460-461

ExposomeExposome--wide association wide association ppstudies (EWAS)studies (EWAS)

By applying EWAS to biospecimens from healthy and diseased subjects, we can discover causal environmental exposures.

http://www.flickr.com/photos/paulieparker/246707763/

But which ‘omes’ offer the most promise f S f ?

16SM Rappaport

for EWAS and follow-up studies?S. Rappaport, Biomarkers, 2012, 17(6), 48: 3-9

The molecular basis of life The molecular basis of life (and disease)(and disease)

Genome Genome (G = DNA)(G = DNA)

Transcriptome Transcriptome (R = RNA)(R = RNA)

ProteomeProteome(P = large (P = large

molecules)molecules)

MetabolomeMetabolome(M = small (M = small molecules)molecules)

INTERNAL CHEMICAL ENVIRONMENTINTERNAL CHEMICAL ENVIRONMENT

Disease pathwaysDisease pathwaysCausal pathway (c) Reactive pathway (r)

Disease pathwaysDisease pathways

G R Disease

p y ( )

SecondaryG RcDisease

phenotype RrE

Secondary phenotype

G = genomeE = environmentR = transcriptome (gene expression)R = transcriptome (gene expression)

S. Rappaport, Biomarkers, 2012, 17(6), 48: 3-9Based on: E. Shadt et al., Nat Gen, 2005, 37: 710-717

Adding omesAdding omes

M

Causal pathway (c) Reactive pathway (r)

Disease Secondary

Mr

G Rc PcDisease

traits Rr PrSecondary

traits

E

G = genome

E Mc

gE = environmentR = transcriptome (gene expression)P = proteome (protein expression)M = metabolome (all small molecules and metals) S. Rappaport, Biomarkers, 2012, 17(6), 48: 3-9

More omic connectionsMore omic connections

M

Disease Secondary

Mr

G Rc PcDisease

traits Rr PrSecondary

traits

EMc

Genetic modifications (mutations)

Epigenetic modificationsPost-translational modificationsGenetic modifications (mutations)

S. Rappaport, Biomarkers, 2012, 17(6), 48: 3-9

Which omes for EWAS?Which omes for EWAS?

G Rc PcDisease

traits

EMc

If causal exposures operate primarily through small molecules (M ) and proteins (P ) then EWAS requiremolecules (Mc) and proteins (Pc), then EWAS require metabolomics and/or proteomics.

21SM Rappaport

Biospecimens for EWAS?Biospecimens for EWAS?Reactive biomarkers

(disease)Causal biomarkers

(exposure)( p )Mr

G Rc PcDisease

traits Rr Pr

EMc

Reactive biomarkers obscure causal pathways. For validation of exposure biomarkers, biospecimens should be obtained prior tobiospecimens should be obtained prior to disease (prospective cohorts)

Bioactive moleculesBioactive moleculesM t b lM t b lR ti l t hilR ti l t hil

Bioactive moleculesBioactive moleculesMetabolome:Metabolome:

LipidsLipidsSugarsSugars

NucleotidesNucleotides

Reactive electrophiles:Reactive electrophiles:Reactive O, N & Cl speciesReactive O, N & Cl species

AldehydesAldehydesEpoxidesEpoxides

Inflammation markers:Inflammation markers:CytokinesCytokines

ChemokinesChemokinesEicosanoidsEicosanoidsAmino acidsAmino acids

MetabolitesMetabolitesXenobioticsXenobiotics

QuinonesQuinones EicosanoidsEicosanoidsVasoactiveVasoactive aminesamines

Growth factorsGrowth factors

SERUM EXPOSOMESERUM EXPOSOMEMetalsMetals

MicronutrientsMicronutrients ReceptorReceptor--binding agents:binding agents:HormonesHormones

XenoestrogensXenoestrogens

SERUM EXPOSOMESERUM EXPOSOMEXenoestrogensXenoestrogens

Endocrine disruptorsEndocrine disruptorsMicrobiome Microbiome productsproducts DrugsDrugs

S. Rappaport, Biomarkers, 2012, 17(6), 48: 3-9

Diseased vs. healthy ( t l t di )

Serum exposome(case-control studies)Untargeted designs

Discriminating featuresChemical

Candidate biomarkers

DATA-DRIVEN DISCOVERY (EWAS)

Chemical identification

Biomarkers of exposure

Diseased vs. healthy (prospective cohorts)Targeted designs

Biomarkers of disease

S. Rappaport, Biomarkers, 2012, 17(6), 48: 3-9

Diseased vs. healthy ( t l t di )

Serum exposome(case-control studies)Untargeted designs

Discriminating featuresChemicalDATA-DRIVEN

DISCOVERY (EWAS) Candidate biomarkers

Chemical identification

Biomarkers of exposure

Diseased vs. healthy (prospective cohorts)Targeted designs

Biomarkers of disease

Molecular Systems

Genomics , epigenomics, transcriptomics & experiments

Identify sources & measure exposures

Exposure

KNOWLEDGEKNOWLEDGE--DRIVEN DRIVEN APPLICATIONSAPPLICATIONS

Drug

Disease stage and response to therapy

Molecular epidemiology

Systems biology

Exposure biology

Drug development

Causality andprevention

Diagnosis, prognosisand treatmentS. Rappaport, Biomarkers, 2012,

17(6), 48: 3-9

Diseased vs. healthy ( t l t di )

Serum exposome(case-control studies)Untargeted designs

Discriminating featuresChemicalDATA-DRIVEN

DISCOVERY (EWAS) Candidate biomarkers

Chemical identification

Biomarkers of exposure

Diseased vs. healthy (prospective cohorts)Targeted designs

Biomarkers of disease

Molecular Systems

Genomics , epigenomics, transcriptomics & experiments

Identify sources & measure exposures

Exposure

KNOWLEDGEKNOWLEDGE--DRIVEN DRIVEN APPLICATIONSAPPLICATIONS

Drug

Disease stage and response to therapy

Molecular epidemiology

Systems biology

Exposure biology

Drug development

Causality andprevention

Diagnosis, prognosisand treatmentS. Rappaport, Biomarkers, 2012,

17(6), 48: 3-9

EWAS: proof of conceptEWAS: proof of concept(M t b l i i NMR & MS)(M t b l i i NMR & MS)(Metabolomics via NMR & MS)(Metabolomics via NMR & MS)

Candidate Modest numbers

27

biomarkersodest u be sof subjects

S. Rappaport, Biomarkers, 2012, 17(6), 48: 3-9

An EWAS of colorectal An EWAS of colorectal cancercancer

Possible omic features:

28SM Rappaport Ritchie et al., BMC Medicine, 2010, 8, 13

Possible omic features: 900 Da x 500 features/Da ≈ 0.5M features

Biomarker identificationBiomarker identificationBiomarker identificationBiomarker identification•• Structures not confirmedStructures not confirmed

•• Hydroxylated ultraHydroxylated ultra--longlong--chain chain fatty acids (Cfatty acids (C2828 –– CC3636) )

•• UniqueUnique--mass spectra permitmass spectra permitUniqueUnique mass spectra permit mass spectra permit precise measurementsprecise measurements

•• Probably antiProbably anti--inflammatory inflammatory agents similar to resolvinsagents similar to resolvinsagents similar to resolvins, agents similar to resolvins, protectins and protectins and lipoxinslipoxins(products of omega(products of omega--3 fatty 3 fatty acids)acids)

Resolvin E1

SM Rappaport29

Follow up measurements of CRCFollow up measurements of CRC--446446Does not return to normal after treatment

Uncorrelated with CRC stage

Biomarker highly associated with CRC Biomarker also decreases with age

Results indicate that CRC-446 may be a

30SM Rappaport Ritchie et al., BMC Gastroenterology, 2010, 10, 140

ycausal biomarker of (protective) exposure!

Two biomarkerTwo biomarker--research agendasresearch agendasEWASEWAS

oo For disease etiologyFor disease etiologyoo DataData--driven, untargeted designsdriven, untargeted designsoo Focus on small molecules and proteinsFocus on small molecules and proteins

T id tif bi kT id tif bi koo To identify biomarkersTo identify biomarkersoo Proof of concept has been establishedProof of concept has been established

FollowFollow up studiesup studiesFollowFollow--up studiesup studiesoo For etiology, diagnosis and prognosisFor etiology, diagnosis and prognosisoo KnowledgeKnowledge driven targeted designsdriven targeted designsoo KnowledgeKnowledge--driven, targeted designsdriven, targeted designsoo For causative or suspicious factorsFor causative or suspicious factorsoo Use biomarkers to confirm causality, etc.Use biomarkers to confirm causality, etc.oo Use biomarkers to confirm causality, etc.Use biomarkers to confirm causality, etc.oo Provide feedback for public health and treatmentProvide feedback for public health and treatment

31SM Rappaport

Needs for EWAS and followNeeds for EWAS and follow--upup1.1. Interdisciplinary research teams (e.g. epidemiology, Interdisciplinary research teams (e.g. epidemiology,

medicine, toxicology, analytical chemistry and medicine, toxicology, analytical chemistry and i i /bi i f i )i i /bi i f i )statistics/bioinformatics) statistics/bioinformatics)

2.2. Apply untargeted omics (metabolomics, proteomics Apply untargeted omics (metabolomics, proteomics dd dd t idd t i ) t lti l) t lti l t l t dit l t diand and adductomicsadductomics) to multiple case) to multiple case--control studiescontrol studies

oo StateState--ofof--thethe--art equipment (HRart equipment (HR--MS/MS)MS/MS)Method development/validationMethod development/validationoo Method development/validationMethod development/validation

oo Identify discriminating features (candidate biomarkers)Identify discriminating features (candidate biomarkers)33 Follow up with biospecimens from prospectiveFollow up with biospecimens from prospective--cohortcohort3.3. Follow up with biospecimens from prospectiveFollow up with biospecimens from prospective--cohort cohort

studies (targeted designs)studies (targeted designs)oo Add transcriptomics and systems biologyAdd transcriptomics and systems biologyp y gyp y gyoo Advanced bioinformatics and statisticsAdvanced bioinformatics and statistics

32SM Rappaport

Best wishes from Berkeley

Major support from NIEHS through grants U54ES016115 and P42ES04705Major support from NIEHS through grants U54ES016115 and P42ES04705